Results

Analysis of lymphocyte proliferative responses to PHA and PWM did not show any significant differences between patients and controls, indicating that non-specific T cell mediated immune responses were not affected in patients (data not shown).

Figure 1 shows the stimulation indexes of lymphocyte proliferative responses to drugs in patients and controls. For each patient, the stimulation index represented corresponds to the higher lymphocyte response observed in the five different concentrations of the drug tested. Part A of the figure shows the stimulation indexes observed in standard conditions—that is, lymphocyte cultures with drugs alone. Significant proliferative responses to drugs (SI>2) were seen in 25 (26%) patients with a clinical diagnosis of drug induced liver injury. In these positive cases, stimulation indexes ranged from 2.1 to 41.2. The four cases with higher stimulation indexes correspond to cases of pyritinol induced prolonged cholestatic hepatitis. It is worth noting that no significant reactivity was found in the three types of controls studied. Part B of the figure shows the stimulation indexes observed when a prostaglandin inhibitor was added to the lymphocyte cultures. Positive results (SI>2) were observed in 53 (56%) patients with a clinical diagnosis of drug induced liver injury. Thus, a significant number of positive cases (n=28) was found only after use of this method. In most cases, an increase in the stimulation indexes was observed when a prostaglandin inhibitor was added to the cultures, even in cases already positive in standard conditions. A notable exception to this phenomenon was observed in the group of non-steroidal anti-inflammatory drugs (NSAIDs).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

: Lymphocyte stimulation indexes in patients and controls (A) in the presence of the suspected drug and (B) in the presence of the suspected drug and a prostaglandin inhibitor. The numbers in brackets represent the stimulation indexes of cases that fall outside the scale. The horizontal line represents the level of stimulation index above which results are considered positive.

Table 5 lists the drugs to which specific T cell reactivity was detected. The total number of drugs studied was 152, corresponding to a mean of 1.6 drugs per patient, with limits between 1 and 6. In all cases there was agreement between the drugs identified by immunological studies and the drugs suspected after a standard clinical imputation process was applied.28 In cases in which two or more drugs were equally suspected by clinical assessment, lymphocyte reactivity was found to one drug only. In three cases of cotrimoxazole induced hepatitis, lymphocyte reactivity was tested both to sulphamethoxazole and trimethoprim, and in all cases reactivity was observed only to sulphamethoxazole. In another case of erythromycin induced cholestatic hepatitis, in which the patient had been exposed successively to different salts of erythromycin, lymphocyte reactivity was tested against ethylsuccinate, estolate, and lactobionate, as well as to base erythromycin; lymphocyte sensitisation was detected only to lactobionate.

In five cases, lymphocyte proliferative responses to the suspected drug were detected on two different occasions (table 6). In three cases the reactivity decreased and in two no changes were observed.

Table 7 presents the possible associations between the presence of PPSC in vitro and the clinical characteristics of patients. Statistically significant differences were found between cases with and without evidence of PPSC when considering duration of latency (p=0.014) and remission (p=0.013), severity of hepatitis as assessed by alanine aminotransferase (p=0.027) and aspartate aminotransferase (p=0.012) concentrations, and the pharmacological groups of the drugs inducing hepatitis (p=0.017).

Cases with evidence of PPSC had a significantly lower median latency period (11 days) than cases without PPSC (15 days) as assessed by the Mann-Whitney U test (p=0.042). The median duration of remission was also lower in the first group (43 v 65 days) although this did not reach statistical significance (p=0.146).

Table 8 shows the global results of specific T cell reactivity to drugs tested in 95 patients with a clinical diagnosis of drug induced liver injury, using different methods. It is worth noting that the addition of a prostaglandin inhibitor to the lymphocyte cultures increased significantly the ability to detect lymphocyte reactivity to drugs. The results obtained in a series of 25 cases previously published18 in which ex vivo drug antigens were used instead of the parental drug are also shown in table 8 for comparison. Although the percentage of positive cases with this approach was only 28%, several of these cases were not detected when the test was done in standard conditions or with simple addition of a prostaglandin inhibitor. Finally, in cases where the three different approaches were simultaneously applied—that is, lymphocyte proliferation to drugs alone, lymphocyte proliferation to drugs in the presence of a prostaglandin inhibitor, and lymphocyte proliferation to ex vivo drug antigens, a much higher percentage of positive cases was observed (88%).

In order to assess whether a positive in vitro test was associated with the presence of extrahepatic manifestations which are usually considered to be a clinical marker of an immunological mechanism for drug hepatotoxicity, we analysed the frequency of these extrahepatic manifestations in patients with positive and negative test results. No significant differences were found, as assessed by Fisher’s exact test.

Discussion

Specific T cell reactivity to drugs has been used in an attempt to improve diagnosis of drug induced liver injury but a low sensitivity of the test has been reported.14 ,15 Most studies investigated individual cases or small series. We have analysed the value of this method in a large series of well characterised patients, subject to long follow up periods. We used modifications to the standard tests, in an attempt to overcome some of the problems that could theoretically explain the low sensitivity of the test—namely, the presence of prostaglandin producing suppressor cells in the in vitro culture system.17 ,23 ,29 Our results showed that this modification increased significantly the sensitivity of the test, without decreasing its specificity. In fact, no reactivity was found in any of the three types of controls studied. Control subjects with recent exposure to the suspected drugs without adverse effects are particularly important in this kind of study, because sensitisation to the drug could be an epiphenomenon observed in those exposed to the drug and unrelated to the hepatic lesion. Conversely, some liver diseases could somehow alter the immune system and predispose to non-specific sensitisation to drugs. Negative results observed in our controls with liver diseases unrelated to drugs exclude this possibility.

The use of a large number of concentrations of the drug seems to be of considerable importance, because in several cases lymphocyte reactivity was detected only to one or two concentrations of the drug. This finding is not totally unexpected as the optimal concentration of the drug for generating the adequate antigenic presentation in vitro is unknown. Furthermore, it is worth noting that the optimal concentration of a certain drug varies from one person to another. Therefore, the criteria of choosing a concentration corresponding to the average plasma therapeutic concentration of the drug and two concentrations above and two below that level are clearly arbitrary, although it was shown to be of practical importance in the sense that an optimal concentration could be either below or above the chosen reference concentration.

As suggested in our early studies,16 ,17 the increase in lymphocyte proliferation in response to drugs by the addition of a prostaglandin inhibitor to the cultures suggests that it is the presence of prostaglandin producing suppressor cells that prevents lymphocyte sensitisation to drugs being detected in standard conditions. This in vitro phenomenon seems to be clinically relevant in vivo, in the sense that patients in which this phenomenon was detected had a less severe disease, as assessed by aminotransferase concentrations and more rapid cure. This observation is interesting in view of the data showing a protective role of prostaglandins, namely PGE₂, in experimental models of liver injury,30 as well as in patients with fulminant hepatic failure.31 Recent studies have shown that PGE₂profoundly inhibits the synthesis of interleukin 2 and interferon γ,32 ,33 two lymphokines associated with the Th1 pattern, but not of interleukin 4 and interleukin 10 secreted by Th2 cells. There is growing evidence that the Th1 subset may be involved in organ specific autoimmune diseases, not only in animal models,34 but also in humans, as is the case in multiple sclerosis35 and rheumatoid arthritis.36 Many cases of drug induced prolonged cholestasis associated with the vanishing bile duct syndrome show some features in common with autoimmune liver diseases, such as primary biliary cirrhosis (PBC).37 ,38 It has been shown that biliary epithelial cells from patients with PBC exhibit aberrant expression of HLA class II antigens39 in association with T cell autoreactivity against biliary autoantigens,40 and this abnormal expression could be dependent on interferon γ.41 ,42Thus, it is tempting to speculate that, by inhibiting interferon γ and favouring a switch from the Th1 to the Th2 subset, PGE₂may have an important role in preventing drug induced autoimmune liver disease.

An additional factor that could contribute to the high rate of detection of sensitised T cells in our series may be related to the rigorous and strict criteria we used in establishing the diagnosis of drug induced liver injury. Furthermore, we believe that in our series cases of liver damage induced by dose independent idiosyncratic metabolic mechanisms are under-represented.

Antihypertensives and some antimicrobial drugs seem to be more prone to inducing prostaglandin producing suppressor cells than other pharmacological groups. The inability to detect this phenomenon in the NSAID group is probably because these drugs are already prostaglandin inhibitors and therefore no additional effect is obtained with the addition of indomethacin.

We do not question the usual practice of diagnosing drug induced liver injury by clinical criteria, with emphasis on the exclusion of alternative causes and the establishment of a consistent temporal relation between drug intake and the onset of clinical manifestations.2 However, in patients exposed to more than one drug and particularly for recently marketed drugs, the demonstration of specific T cell reactivity to a given compound represents additional evidence in favour of its involvement in the induction of liver injury6 and is also important for defining the type of mechanism involved.

It is usual clinical practice to assume that the presence of clinical manifestations such as rash, arthralgia, and eosinophilia is the best marker for an immunological mechanism in drug induced hepatotoxicity.4 The fact that such clinical markers were absent in about 70% of our patients with a demonstrable T cell reactivity to drugs challenges this view. Furthermore, our results showing lymphocyte sensitisation to drugs in as many as 65% of the patients suggest that immunological mechanisms are much more frequent in drug hepatotoxicity than previously assumed.7 However, this does not preclude the possibility of coexistence of metabolic idiosyncratic mechanisms that could be responsible for the formation of reactive metabolites involved in the induction of the abnormal immune response.

Cross reactivity between drugs from the same pharmacological group often occurs and it is normally recommended that a patient with hepatitis associated with a given drug should avoid taking any drug belonging to the same group.43 However, our preliminary results on this topic suggest that drug sensitisation seems to be highly specific. In fact, in several cases in which patients had been exposed to pharmacologically related compounds or even to different salts of a given drug, lymphocyte reactivity was detected only to one drug or to a particular drug salt. Furthermore, in the rare instances (five cases) in which patients had been tested more than once within a period of few weeks after the onset of the clinical sysptoms, reactivity was always detected to the same drug. Taken together, these data clearly show the high specificity of these in vitro tests.

The notable increase in the sensitivity of drug specific T cell reactivities obtained with this new approach, together with its high specificity, could lead to a recommendation for the more general use of these tests in clinical practice.44 The only limitation to such widespread use is that the relative complexity of this method implies that such tests should be confined to specialised laboratories with experience in in vitro functional cellular immunology techniques. However, the renewed interest in drug induced liver injury and the importance of more accurate diagnoses probably justifies the effort required of specialised laboratories to make such tests available in the investigation of selected cases of drug induced hepatitis.

In conclusion, modified lymphocyte transformation test used in our study significantly increased our ability to detect lymphocyte sensitisation to drugs without altering the specificity of the test. We therefore suggest that the study of specific T cell reactivity to drugs should have a place in the diagnosis of drug induced liver injury, particularly in patients exposed to more than one drug.